Investigating soil microbial communities - differential abundance (ANCOMBC)
Kerry Hathway
2025-09-01
Differential abundance analysis using ANCOMBC was carried out on unrarefied data due to the recommendations by the package authors and discussions online. For significant taxa associated with different site-treatment groups, the data was non-aggregated to ensure there were enough samples for ANCOM analysis.
Bacteria
Treatment
No taxa at phyla or genera level were identified as differentially abundant when comparing control and water stress samples overall.
We also tested for phyla, families or genera which were differentially abundant between control and drought on individual sites. Again, ANCOMBC did not identify any significant taxa. This was surprising given PERMANOVA tests determined that treatment had a significant effect on community composition for unweighted UniFrac.
A reason for this could be the small sample sizes. Bioinformatics tools such as ANCOMBC are less effective for low sample counts, while small sample sizes can increase the risk of Type II errors. These issues may have hampered our ability to find significant differences between treatments
See further down for results which were not significant
Comparing sites within treatments
However, when we compared sites within treatments, we had some significant results. Multiple phyla and genera passed both ANCOMBC’s differential abundance and sensitivity tests.
For both control and drought, several phyla appeared as differentially abundant between sites in both treatments. For example, Gemmatimonadota, Chloroflexota and Eisenbacteria.
Some phyla were not differentially abundant in control samples but showed strong differences between sites in the drought samples, indicating a drought response.
Below are the results of the global test for controls. Site comparisons (in the tabs) have been split into separate tables containing all comparisons for that particular site.
For our analysis, we only considered taxa which had passed both the differential abundance and sensitivity tests. Taxa which passed one test but not the other were disregarded. However, the results from all taxa are shown in the tables below.
Control - phyla
For the control samples, differences in abundance under ambient conditions indicated significant phyla were responding to strongly to site characteristics.
Phyla showing the strongest differences between sites in the controls (and which passed both tests) were Nitrospirota A and Gemmatimonadota.
Nitrospirota A contain nitrifying taxa, and is a phyla which is generally unresponsive to drought, but does respond to site factors such as land management intensity and soil nitrogen availability.
Harley Farms
Castle Field West
Jemma 6
Jemma 9
Lardon Chase
Windmill Farm
Drought - phyla
For the drought samples, Bacillota A-368345, which was not significantly different between sites’ control samples, showed a very strong change in abundance between sites. Bacillota A-368345, a strain in the phyla formerly known as Firmicutes, are drought resistant. Their ability to withstand or thrive in drought conditions is linked to their capacity to form endospores, which remain dormant in the soil and confer a competitive advantage in droughted-soils
Site pairwise comparisons is where we could see several changes happening. Major changes included:
- Bacillota A: -2.59 log2 fold change between Jemma 6 and Harley Farms
- Bacillota A: 2.43 log2 fold change between Lardon Chase and Jemma 6
- Chloroflexota: 1.5 log2 fold change between Windmill Farm and Lardon Chase
Harley Farms
Castle Field West
Jemma 6
Jemma 9
Lardon Chase
Windmill Farm
Control - genera
Due to time contraints, we were not able to carry out a detailed analysis of changes at a genera level, but there were some significant results for both control and drought samples. Full tables are below
Harley Farms
Castle Field West
Jemma 6
Jemma 9
Lardon Chase
Windmill Farm
Drought - genera
Harley Farms
Castle Field West
Jemma 6
Jemma 9
Lardon Chase
Windmill Farm
Fungi
Treatment
As with bacteria, no differentially abundant phyla associated with treatment, either overall or on individual sites, were identified by ANCOMBC analysis.
Again, this may be due to small sample sizes. The extra challenge we had was missing samples for some sites due to quality control failures. ANCOMBC needs at least three samples per factor and both Windmill Farm and Castle Field West had only two samples in some treatments. This meant, treatment comparisons on these sites could not be carried out.
Comparing sites within treatments
When comparing sites within treatments, we had some significant results although not as many as for bacteria. Additionally, the p-values in the global tests were not as low as for bacteria, indicating a weaker response, particularly to drought conditions.
For control samples, only four phyla were differentially abundant in global tests:
- Glomeromycota
- Rozellomycota
- Kickxellomycota
- Zoopagomycota
Control - phyla
Jemma 6
Jemma 9
Lardon Chase
Windmill Farm
Drought - phyla
For drought samples only one phylum was differentially abundant between sites: Zoopagomycota
Castle Field West
Jemma 6
Jemma 9
Lardon Chase
Control - genera
Time constraints meant we were unable to complete a full analysis of how abundances change for specific genera. The results tables are below.
Jemma 6
Jemma 9
Lardon Chase
Windmill Farm
Drought - genera
Castle Field West
Jemma 6
Jemma 9
Lardon Chase
Non-significant results
Below are the outputs from significance tests for treatment, where there were no significant results. These are included for academic transparency and in case they are of interest.